Franklin Templeton: Agentic AI is the blockchain’s “killer application”

Byline: Sandy Kaul, Head of Digital Assets and Innovation at Franklin Templeton

Translation: Jiahuan, ChainCatcher

AI evolution is driving the investment narrative

AI has been evolving. In the early days around the 2010s, early capabilities such as machine learning, natural language processing, and predictive analytics sparked the “big data” era—enabling people to process structured and unstructured data at speeds and scales that were previously unimaginable. Back then, AI was more like a tool that helped humans at work.

In the early 2020s, generative AI emerged, and its use cases and role took a major leap forward. AI shifted from being an assistant to a “co-creator”: it can generate content, respond to all kinds of questions, and even help complete parts of work for people. The potential of generative AI has not been fully unleashed yet—products are getting stronger, and it is also seeping into more corners of everyday life.

As its influence expands, AI’s position as the investment mainline is now beyond dispute.

On July 14, 2026, IBM’s stock price plunged 25.2% in a single day. Earlier, it issued a warning: companies’ technology budgets are increasingly being directed toward AI infrastructure, while spending on traditional software and IT projects is being delayed or cut.

Today’s S&P 500 index has concentration at its highest level since the late-1990s tech bubble. The top 10 stocks by market cap are all AI-related, together accounting for nearly 40% of the total market cap of the entire index. By comparison, during the internet bubble period this number was only 25%, and in 1980 it was just 15%.

Institutional investors, in particular, view AI as a structural major trend and are heavily betting on AI infrastructure, data centers, and semiconductor stocks.

But this kind of allocation may still be insufficient to capture the upside of AI’s next round of evolution.

The rise of Agent AI

Compared with earlier AI tools, generative AI at the time was a leap in capabilities, outcomes, and application scenarios. Today, as Agent AI matures and is widely adopted, its impact on daily life may be no less than—or even greater than—what came before.

Agent AI moves the interaction model from a passive, conversational chatbot to an autonomous system that can sense its environment, make its own plans, and execute multi-step tasks to achieve high-level goals without people continuously watching.

Agents can directly interact with external software systems and code repositories—so the role of AI is being redefined. 38% of institutions say that by 2028, Agents will become team members like human colleagues, working together to improve productivity and drive innovation.

Following this trend, the tasks assigned to AI will become increasingly complex. Generative AI is good at gathering knowledge and organizing content; Agents will increasingly take on “transaction-like” work—initiating, tracking, and completing tasks themselves, and managing the final results.

Some predictions suggest that by 2030, the commercial scale of Agent business could reach as high as $3 trillion to $5 trillion.

For institutions, most of these transactions will happen inside enterprise software. Predictions show that by 2028, 33% of enterprise software will embed Agent AI, and up to 15% of daily decisions will be handled by these Agents.

Between software and software, they will pay each other small fees for compute power, API calls, data usage, and various services. This is a brand-new interaction model, allowing precise bookkeeping and settlement on a per-task basis.

Protocols for “software paying software” are emerging

Protocols that support these “machine-to-machine” transactions are rolling out one after another. Stripe and Visa have already launched a machine payment protocol (MPP).

Open-source solutions are also gaining traction. In the early 1990s, when the designers of the World Wide Web were setting rules for browser-to-server communication, they specifically reserved an HTTP response code numbered “402,” labeled “Payment Required.”

Coinbase built an “x402” protocol on top of this, allowing Agents to initiate and complete payment instructions of this kind, and then it further handed over the related intellectual property to the Linux Foundation, making it an open industry standard.

Today, major credit card networks, Web2 giants like Stripe, Shopify, Google, and Amazon Web Services (AWS), and an increasing number of Web3 service providers have already integrated this payment standard. The goal is to enable “software paying software,” with no human involvement needed end to end.

In the next few years, Agent payments will very likely reshape how consumers interact. Some predictions say that by 2030, Agents will contribute 15% to 25% of U.S. e-commerce sales.

Right now, ChatGPT handles 2.5 billion questions per day, and shopping-related queries initiated via AI platforms account for 53 million times. OpenAI is also moving checkout into third-party ChatGPT apps, such as Target, DoorDash, and Instacart.

Blockchain: how to support machine-to-machine transactions

To enable these “machine-to-machine” transactions, you need a secure, autonomous, verifiable, high-throughput accounting system.

Traditional credit card and banking systems are not suitable for Agent’s small payments in terms of fee structure. A standard credit card transaction typically charges a 2% to 3% fee, plus a fixed cost of about $0.3. Meanwhile, when an Agent buys 1 second of compute or executes 1 data query, the average cost is only $0.001.

To get Agent AI running, it will likely rely on cryptography and blockchain, because this underlying rails are naturally well-suited for such scenarios. In fact, thanks to the following features, blockchain and cryptography are poised to become the underlying layer for these transactions.

Automatic generation and execution of contracts. Payment Agents generate tokens to complete purchases and settlements. Each token includes a set of transaction rules: which merchants can accept the token, the maximum amount that can be spent per transaction, and how long the token remains valid. Once a purchase is completed, such a one-time token automatically becomes invalid. Blockchain can hold, send, and receive these tokens, and—just like executing smart contracts—strictly enforce the rules written inside the tokens.

Decentralized identity verification. Each Agent has a unique identity that can be verified using cryptography. Every token it generates carries its own credentials; only with those credentials can the Agent sign blockchain transactions. When validating transactions, the blockchain checks these credentials. If the identity is deemed invalid, the consensus mechanism will block the transaction.

Full auditability. Every decision an Agent makes on-chain, every transaction, and every data exchange can be recorded on an immutable ledger. Anyone can publicly retrieve it via a blockchain explorer, ensuring accountability and transparency throughout the process.

Access to decentralized compute power and data. With blockchain, Agents can call distributed compute resources (such as GPU networks) and data, reducing reliance on centralized, private-cloud infrastructure, and also helping lower the operating costs of high-frequency transaction models.

Speed and settlement. Bitcoin can process about 7 transactions per second, Ethereum about 75, but newer high-speed public chains have pushed peak throughput higher: Aptos can reach up to 12,933 TPS, Solana 6,284 TPS, and BNB Chain 3,252 TPS.

This speed is already comparable to the Visa network, which processes 1,700 to 10,000 transactions per second during normal operation. But even so, this still underestimates on-chain systems. In the TPS time window, blockchain both records and completes settlement; Visa only records the transaction, and real settlement still takes 1 to 3 business days.

With these characteristics, blockchain will play a key role as Agent AI rolls out into consumer-grade transactions. Conversely, the growth of Agent AI may also become the “killer application” that drives mainstream adoption of blockchain.

How to invest in the Agent AI opportunity

At present, to capture the upside from AI growth, investors typically buy stocks of AI concept companies and companies in related industrial chains, or they participate via LPs in private equity funds, or they invest in energy providers and data centers that support AI operations.

But to seize the opportunity in Agent AI, these approaches may still need to extend exposure to native tokens of public chains, as well as the project tokens issued by on-chain applications and projects. Several forces are likely driving this shift.

Demand for cryptocurrencies will rise. To record a transaction on a given chain, an Agent must pay transaction fees using that chain’s native token. For example, to record on Solana, you need to pay SOL. As Agents make more payments, the demand for native tokens supporting these businesses could surge, creating value for every token holder. At first, this demand will most likely come from machine-to-machine micro-payments between enterprise software systems.

The blockchain ecosystem will expand. The more transactions on a chain and the stronger the native token demand, the more money will flow into that chain’s treasury. Blockchain foundations will use these treasury funds to grant developers for building, encourage them to develop applications on-chain, pay “bug bounties” to developers who discover security vulnerabilities, and also incentivize those who verify transactions for the network. The more money there is to distribute, the more likely the ecosystem will grow—becoming increasingly secure—thereby attracting more developers to build applications, issue their own tokens to finance projects, and share ownership of the applications.

Web3 applications will take share from Web2. As more applications move on-chain and more development talent flows in, Web3 applications’ advantages over Web2 will become increasingly obvious. This has already happened in Web3 gaming: the industry is shifting from Web2’s “single-player” model to Web3’s “players own the economy.”

Tap-to-earn (T2E) apps have already attracted hundreds of millions of users worldwide. Players can now truly own and monetize the assets they accumulate in games via the secondary market, cross-platform trading, and buying and selling in-game items (NFTs). Similar ownership-and-experience changes may play out across a large number of consumer-grade applications, which would also boost market interest in the tokens issued by these projects.

The flywheel effect will kick in. Protocols that embed Agent payments into blockchain applications already exist, and they may create a flywheel effect for the newly issued tokens. Users only need to instruct their Agents to handle transactions and payments; they don’t have to create wallets themselves, buy tokens, manage various tokens and cryptocurrencies.

For users, the experience of using a Web3 app may feel no different from using a Web2 product. But because the tokens here have real utility value and also represent ownership, users can get more favorable outcomes within the Web3 ecosystem.

To some extent, this transition resembles the earlier shift from Web1 to Web2: from Web1’s webpage servers and static websites to Web2’s cloud services and interactive applications. In both transitions, whether you look at the providers of underlying technology or the businesses built on top of those rails, the protagonists shifted from the original established players to a new batch of players that drive growth in the新时代. This time, the baton will be passed to blockchain and to various decentralized applications and projects.

Right now, investors still haven’t fully figured out how to capture the value created by blockchain and its ecosystem. They’re used to a centralized, company-led business world: if they want to share in a company’s created value, they just buy its stock.

But I believe what will become increasingly clear over the next few years is this: to capture the value of decentralized networks and businesses, investors need to buy relevant crypto assets. These assets are likely to become important holdings in investment portfolios—and especially so for those looking to catch the new opportunity in Agent AI.